Keep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation

Not provided

text-to-imagesafetycounterfactualAI ethics

2610.02300

Builder Relevance

80%

1h ago

Abstract

The paper analyzes the limitations of global unsafety in text-to-image generation and proposes a new safeguard method called CALM.

Reality Card

Core Claim

CALM significantly improves unsafe content suppression while preserving benign utility by using prompt-local counterfactual correction instead of global unsafe signal removal.

Method / Result

CALM demonstrates improved unsafe content suppression while maintaining benign utility across broad evaluation.

Limitations

The paper does not specify the reproducibility of the results or the generalizability of CALM across different models or datasets.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers